[Core] throw error when patch inputs and layernorm are provided for Transformers2D

[sayakpaul]

What does this PR do?

Fixes: #7079 by throwing an error.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[yiyixuxu]

can we also add this PR to update the signature for norm_type to include everything we support for norm_type. we don't have signature here so maybe just add a comment like we did here

diffusers/src/diffusers/models/attention.py

Line 146 in fa1bdce

norm_type: str = "layer_norm", # 'layer_norm', 'ada_norm', 'ada_norm_zero', 'ada_norm_single', 'layer_norm_i2vgen'

we accept all of below, but I think we only had model that use patch with ada_norm_zero and ada_norm_single, no? can we double check which norm_types can be used with patch_size and throw and error accordingly (right now in the PR we assume that everything other than layer_norm is supported)

        `layer_norm', 'ada_norm', 'ada_norm_zero', 'ada_norm_single', 'layer_norm_i2vgen'

[yiyixuxu]

thanks!